Research map / Vision-language and multimodal / Cross-modal learning
RingMoE: Mixture-of-Modality-Experts Multi-Modal Foundation Models for Universal Remote Sensing Image Interpretation
Hanbo Bi, Yingchao Feng, Bowei Tong, Mengyu Wang, Haichen Yu, Yongqiang Mao, Hao Chang, Wenhui Diao, Peijin Wang, Yue Yu, Hanyang Peng, Yehong Zhang and 2 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add modal-specialized experts (ablation) | RingMoE · NWPU-RESISC45 optical classification | accuracy: 93.78 -> 95.05 | Improved |
| add modal-specialized experts (ablation) | RingMoE · Dyna.-pla multi-spectral segmentation | mIoU: 46.28 -> 47.60 | Improved |
| add collaborative experts (ablation) | RingMoE · WHU-OPT-SAR multi-modal segmentation | mIoU: 52.63 -> 54.70 | Improved |
| add shared expert (ablation) | RingMoE · NWPU-RESISC45 optical classification | accuracy: 94.59 -> 95.05 | Improved |
| add physics-informed SAR-L1 reconstruction loss instead of pixel-wise loss (ablation) | RingMoE · SARSegL1 | mIoU: -5.67% mIoU | Improved |
| add physics-informed SAR-L1 reconstruction loss instead of pixel-wise loss (ablation) | RingMoE · SAR-AIRCraft | mAP: -3.14% mAP | Improved |
| scale backbone from 88M to 1B with RMoE layers (ablation) | RingMoE · NWPU-RESISC45 | OA: 93.64 -> 95.90 | Improved |
| scale backbone from 88M to 1B with RMoE layers (ablation) | RingMoE · AID | OA: 96.57 -> 98.19 | Improved |
| prune experts to top 25% instead of keeping all (ablation) | RingMoE-Opt · NWPU-RESISC45 | accuracy: 95.90 -> 95.43 | No clear effect |
| use percentile top-25% pruning instead of fixed threshold 0.1 (ablation) | RingMoE-Opt · NWPU-RESISC45 | accuracy: 95.06 -> 95.43 | Improved |
| retain more experts (top 75% instead of top 25%) (ablation) | RingMoE-Opt · NWPU-RESISC45 | accuracy: 95.43 -> 95.71 | Improved |
| compress experts by SVD instead of summing weights into one FFN (ablation) | RingMoE (1B) · NWPU-RESISC45 | OA: 94.81 -> 95.05 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SkySense (2B) — SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery (CVPR 2024) · better on 5 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery (CVPR 2024 · similarity 0.91)
- Galileo: Learning Global & Local Features of Many Remote Sensing Modalities (ICML 2025 · similarity 0.87)
- SkyMoE: A Vision-Language Foundation Model for Enhancing Geospatial Interpretation with Mixture of Experts (AAAI 2026 · similarity 0.87)
- RobSense: A Robust Multi-modal Foundation Model for Remote Sensing with Static, Temporal, and Incomplete Data Adaptability (CVPR 2025 · similarity 0.87)
- GeoLink: Empowering Remote Sensing Foundation Model with OpenStreetMap Data (NeurIPS 2025 · similarity 0.87)
- Bridging Remote Sensors with Multisensor Geospatial Foundation Models (CVPR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
